1 / 90

Home-grown Crypto (Taking a Shiv to a Gun Fight)

Home-grown Crypto (Taking a Shiv to a Gun Fight). March 25, 2007 Hank Leininger <hlein@korelogic.com> F980 A584 5175 1996 DD7E C47B 1A71 105C CB44 CBF8 Klayton Monroe Cofounders, KoreLogic. Agenda. Intro Case Studies - Home Grown Crypto Wall of Shame Prehash Analysis

Download Presentation

Home-grown Crypto (Taking a Shiv to a Gun Fight)

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Home-grown Crypto(Taking a Shiv to a Gun Fight) March 25, 2007 Hank Leininger <hlein@korelogic.com> F980 A584 5175 1996 DD7E C47B 1A71 105C CB44 CBF8 Klayton Monroe Cofounders, KoreLogic

  2. Agenda • Intro • Case Studies - Home Grown Crypto Wall of Shame • Prehash Analysis • Recommendations and Closing Remarks

  3. Intro • We are not crypto-mathematicians, we're just hackers. • We regularly see bad crypto. • So will you... • If we can break it, crypto-mathematicians can break it before breakfast. • We keep seeing these mistakes, so the message bears repeating.

  4. Intro (2) What kinds of things do we see? • Horrible home-grown “encryption algorithms” • Giving away the key: • For free • For a bag of Cheetos • For a few bucks • Using good, industry-standard algorithms incorrectly • Incorrect assumptions about the work factor required to attack the system

  5. Intro (3) • “Home grown crypto is bad crypto. ... Every programmer tries to build their own encryption algorithm at some point. In one word: Don't.” Brian Hatch, 2003-01-23 • Has everybody heard “Security through obscurity is no security?” • And yet, most of the bad crypto we see boils down to security through obscurity, at best.

  6. Intro (4) What uses of encryption are we talking about? • We're not picking on industry-standard algorithms or protocols such as 3DES, AES, or SSL per se ... • But we will show you cases where we broke implementations protected by them. • We're talking about using encryption for protecting private data, securing communications, generating session IDs, etc.

  7. Agenda • Intro • Case Studies - Home Grown Crypto Wall of Shame • Prehash Analysis • Recommendations and Closing Remarks

  8. Case Study Format • Description of system or application tested • Description of flaw or flaws found • May include screen shots, graphics, code, etc. • How we discovered, broke, or reversed it • How the flaw could have been prevented and/or how we recommended fixing it • But only if “fixing it” was a viable option

  9. Obfuscation Gone Bad • Online Banking Application (Skip)

  10. Obfuscation Gone Bad (1) • Online Banking Application • Awarded a seal of approval from a security company that “certifies” security • Encrypted account numbers were used to • Access account information • Download bank statements and cancelled checks • Unfortunately, the account numbers weren't really encrypted – they were merely obfuscated with a simple substitution cipher.

  11. GET /checkimages/GetImage.asp?FNAME=TGXPPGGG0001760000250 217XRYYYYY&SIDE=FT HTTP/1.0 Referer: https://somebank.owned.com/checkkimages/retrieve .asp?TGXPPGGG0001760000250217XRYYYYY Connection: Keep-Alive Host: somebank.example.com [snip] Cookie: ASPSESSIONIDGQQGQQZF=PEIAHJKCGBHNKJIDBEBKNFLG Obfuscation Gone Bad (2) • Notice the FNAME value in the following request: • TGXPPGGG0001760000250217XRYYYYY • It's a sanitized version of the obfuscated account number ...

  12. Sub DeEncrypt(ThisStr,StrLen) OldStr = ThisStr ThisStr= "" For i = 1 to StrLen tmpChar = Mid(OldStr,i,1) Select Case tmpChar Case "G" ThisStr = ThisStr & "0" Case "X" ThisStr = ThisStr & "1" Case "P" ThisStr = ThisStr & "2" Case "K" ThisStr = ThisStr & "3" Case "Z" ThisStr = ThisStr & "4" Case "L" ThisStr = ThisStr & "5" Case "R" ThisStr = ThisStr & "6" Case "Y" ThisStr = ThisStr & "7" Case "C" ThisStr = ThisStr & "8" Case "T" ThisStr = ThisStr & "9" [snip] Obfuscation Gone Bad (3) • The DeEncrypt() routine reveals the ugly truth! • Code was obtained from the webserver via a view source bug

  13. TGXPPGGG 0001760000250217 XRYYYYY 90122000 0001760000250217 1677777 Obfuscation Gone Bad (4) • Is this your account number? • This scheme made it very easy for us to brute force account numbers and access customer statements and cancelled checks

  14. Keys? We Don't Need No Stinking Keys • Home-grown Single Sign On (SSO) Application (Back)(Skip)

  15. Keys? We Don't Need No Stinking Keys • Home-grown Single Sign On (SSO) Application • SSO between multiple distributed, un-trusting web servers • Usernames and passwords were encrypted and sent as cookies to other sites (in the same DNS domain) • Each remote site would decrypt the cookie and pre-populate the login form with the decrypted username • The initial login was SSL’ed, but the rest of the applications were not, for “performance reasons”

  16. Keys? We Don't Need No Stinking Keys (2) • The implementation made knowledge of the crypto and encryption keys unnecessary • Decrypting any encrypted blob was simply a matter of: • Constructing a GET request with the target ciphertext added to a special cookie • Submit the request • Wait for the server's response • View the HTML to recover the plaintext

  17. GET /loginform.jsp HTTP/1.0 Cookie: REMEMBERUSERNAME=cb0d2cf4d49d68cf21634f45a7ec7a1a LOGIN_USERID=foo&LOGIN_PASS=q1w2e3r4 <input class='bodytxt' type="text" name="LOGIN_USERID" value='q1w2e3r4'> Set-Cookie: REMEMBERUSERNAME=d47d9e3234c44c3cbc858b43afb1cfb1 Set-Cookie: REMEMBERPASSWORD=cb0d2cf4d49d68cf21634f45a7ec7a1a Keys? We Don't Need No Stinking Keys (3) • A client logs in; POST arguments include: • The server sends to the client: • If we then send the password blob to the server as a username: • The server helpfully decrypts it for us:

  18. Keys? We Don't Need No Stinking Keys (4) • Alone, this flaw was not sufficient to pull off a remote attack (unless you have a nearby victim to sniff). It needed a partner in crime... • One of the components on one of the SSO’ed webservers was a bulletin-board style app • Enter XSS – Attack a user via XSS; their encrypted credentials would be exposed to the attacker • Armed with the ciphertext the attacker could simply issue a modified HTTP request to recover the plaintext

  19. We Don’t Need Logic, We’ve Got Crypto! • A Web Services application with a Java client (Back)(Skip)

  20. We Don’t Need Logic, We’ve Got Crypto! (1) • A Web Services application with a Java client • All transactions were SOAP messages – HTTPS POSTs of XML data • All client messages used SOAP Digital Signatures • The server rejected requests without a valid signature • We thought we were going to need to do some Java disassembly, extract the private key, inject our own, etc. • Turned out it was much, much easier than that

  21. We Don’t Need Logic, We’ve Got Crypto! (2) POST /services/FooService HTTP/1.0 Content-Type: text/xml; charset=utf-8 ... <?xml version="1.0" encoding="UTF-8"?> <soapenv:Envelope ...> <soapenv:Header> <wsse:Security ...> <wsse:UserIsAdmin>no</wsse:UserIsAdmin> <ds:SignedInfo> ... </ds:SignedInfo> <ds:SignatureValue> pRM0+gKSDkpX5v7mWivz78oWDgvGkiOV8vpDNV1VWlVqz2XP68WpydRalI1IYxWzMwtc aOMkjJHmnmRyFBAg92QVe7nUoGIxslMgtEzq+8SKyQu7ldF/xFezMQ0fuf9SjuBCHg== </ds:SignatureValue> • A normal client request might look like:

  22. We Don’t Need Logic, We’ve Got Crypto! (3) <?xml version="1.0" encoding="utf-8"?><soapenv:Envelope ... <soapenv:Body><soapenv:Fault> <faultcode>soapenv:Server.generalException</faultcode> <faultstring>WSDoAllReceiver: security processing failed; nested exception is: org.apache.ws.security.WSSecurityException: The signature verification failed</faultstring> <detail> ... • Try to modify the body, and the server will reject it: • We need to extract the private key the client is using, and sign our own messages. Or do we?

  23. We Don’t Need Logic, We’ve Got Crypto! (4) • Although the server checked for valid signatures on client inputs, the client neither verified SSL certificates nor expected the server's responses to be signed • So... we could forge a message from the server to the client using a standard web-proxy, and then the client would spit it back out to the server at the appropriate time, complete with a valid digital signature.

  24. We Don’t Need Logic, We’ve Got Crypto! (5) • It worked. The server checked that the message had a valid signature – and then blindly trusted the contents! • There was no actual back-end business logic that checked “Hey wait, is user123 really an admin?” • This is the Web Services version of a webapp that checks what a user is allowed to do when drawing a menu, but doesn't check when an action is actually performed. Only the oldest business-logic flaw in the book.

  25. “No Worries, We're Using 3DES” • Network-health monitoring appliance (Back)(Skip)

  26. “No Worries, We're Using 3DES” (1) • Network-health monitoring appliance • Globally distributed and located in customers' internal networks • “Phones home” to vendor to upload data • Documentation claimed that sensitive client information was protected through the use of 3DES encryption • The vendor wanted us to QC their work – kudos to them! • Unfortunately, the 3DES encrypted keys were being given away

  27. “No Worries, We're Using 3DES” (2) • We took a forensic approach: • Cracked open the appliance, imaged the hard drive • Used FTimes (ftimes.sourceforge.net) to scour the image for clues (credentials, keys, code, DB accounts, etc.) • A captured transmission revealed that: • HTTP version 1.0 and ftp were used to POST data to a vendor site • The body of the upload contained binary (3DES?) data that had been encoded as ASCII • So far so good, right?

  28. “No Worries, We're Using 3DES” (3) • We thought so too... until we found the tool that did the encryption: • The key was passed in on the command-line • We traced backwards to find how the key had been generated • Then, we traced forwards to see how the binary data was generated • And that's when we saw the key being stuffed in the payload too! (skip example)

  29. “No Worries, We're Using 3DES” (4) A capture of the host reporting home looked random. Well, sort of… POST /cgi-bin/app/app_httpdld HTTP/1.0 Host: upload.example.com Authorization: Basic YWxseW91cmtleXNhcmU6YmVsb25ndG91cwo= [snip] Content-Length: 2441 Content-Type: multipart/form-data 105860100420976322022480.StatusDownload.rcv _dftgeqaSay ;<CE[DCDCBFh@h3h>h_^cfd^_^^b`^geda``^``bf^h#D1DEECr?G><?14hDB1>C@?BDE[851<D8E[C165\D1Bh?EED213;C5BF5B 8[y@`z6<^g^T` &"XfnEl_U]4>RZ79tL2ei[3Z*nZBz}}FhnAkq$j[_I_\gvzqJ3X)4U0;)j`axBe8 yLCw=xrx"4apa1.;H1VqA]w3lV;"+)+7Yej9c"e_h!]<1:,5a[+FaA;wQC^w{:s"I;,(:6pQ\LU7+L }2X\R>21i4c8XQh<]0T$_ EgzhLYp7Li@R9&z&5R2"3?x8HY2i&#S6R[A1@=XdsvbBIZRn0/#\X'3w jVn{R5`ABR\Hwg{<W@[hCR<si*b2BQE4@dV[VVNg:*XE2B:4:$32Y/VNFar)`|G7|/noDr9}5nD]4} d(d^A2qY$I1!e!fb@bR5<0(9}FU\t<!_f76NY{,BHvidIK8FU &z0LC,/'i:mCmXrX^)m Cr6:UI4] ]wi4v6={9^nyI[re\</BtBCRdH=51s_61z=9:aIn={wx 8;Hol<'Kio9)]DE(0gns +H)/34k3yn[! <[AjBWGL+8;g<Q: }0Ct5}1IIN0,aq13z@v<U&p9ofnh#p.w1r#)vG9e+jGhT3gzR*Gn<)rq5141;s [ snip ]

  30. “No Worries, We're Using 3DES” (5) The encoded stream is also sent via ftp to an upload server. 220 FTP server ready. USER upload2 331 Password required for upload2.. PASS L33tDuD3 230 User upload2 logged in. Access restrictions apply. [snip] STOR 238051003320976321714736.DailyDownload.rcv. 150 Opening BINARY mode data connection for 238051003320976321714736.DailyDownload.rcv. `sfrbdgqS_d^y ;<CE[DCDCBFh@hJh4`h`af^c_^^aa`^geda`_e_beadhr19<Ir?G><?14hDB1>C@?BDE[C165\D1Bh?EED213;C5BF5B 89UvrdN=Y!I<d9r4a<E]SHinL;_$!pJg[*B(sDU|p,X`vF0}/e_s9fU@b[K{DdW=Iqj1B[/q CU2IR xQF0Tz|oror2mL[E&if'V)k!4I(|hVyT3NJsH05s"nj2E/i$;o`[`LcSXV1E6/qQy1'S!0V\SvpYI1 (m,SF{Hnq.rqh#)l!k@,$6k> 1;;Yk+T'h3a4nWp6(2\#J35h_C/*x9 )?q[mUweD"$A9?ZlI^#Q,! IYqZn`3 jW4.|Xe:5G67:z@'.!x|3s0UhK#v!\ ,#=|bINW>d+6"IL2@(c#WQG14Y8h1,z_424`}F. [snip]

  31. “No Worries, We're Using 3DES” (6) #!/usr/bin/perl -wT use strict; $::map = <<'EOMAP'; N {" {# {$ {% {& {' {( {) . . . [snip] EOMAP @::toenc = split(/ /, $::map); my @enc_regex; for (my $i=0; $i<$#::toenc; $i++) { $::toasc{ $::toenc[$i] } = $i; my $token = $::toenc[$i]; $token =~ s/([][?{}.+*()&\$|^\\])/\\$1/g; push(@enc_regex, $token); }; $::regex = join('|', @enc_regex); while(@ARGV and $ARGV[0] =~ /^-([a-zA-Z])/) { my $arg = $1; die "Usage: $0 [-h] [-x] [-r] [-l string]\n" if ($arg eq 'h'); $::func = \&Decode if ($arg eq 'r'); $::hexit = 1 if ($arg eq 'X'); shift; }; $::func = \&Encode unless ($::func); while(<>) { &$::func($_); }; sub Decode { while (m/($::regex)/g) { print chr($::toasc{$1}); }; }; sub Encode { foreach my $char (split(//,$_)) { my $ascii = ord($char); my $mapped = $::toenc[$ascii]; [snip] The ASCII encoding wasn’t anything standard (uuencode, base64, hex-encoding, etc.). But we had their binary. After we made a quick dictionary of all possible byte values, a quick perl equivalent encoder / decoder was produced.

  32. “No Worries, We're Using 3DES” (7) Decoding the headers provided us with clues. The three header lines contain transfer metadata: filename, appliance name, transfer type, various control flags, and oh… the 3DES encryption keys. 359574702420976322006928.DailyDownload.rcv tbae_ofeS_`dy ;<CE[DCDCBFh@h3h>hacgcebe^`b`^geda``^^dg`fhr19<Ir?G><?14hDB1>C@?BDE[C165\D1Bh?EED213;C5BF5B Decoded: 359574702420976322006928.DailyDownload.rcv F4371A87 126K kls-tstsrv:p:c:n:359574702420976322006928:DailyDownload: transport-vault.tar:outbackserver

  33. “No Worries, We're Using 3DES” (8) Example decoding: putting all the pieces together Header lines extracted from a transfer: 238051003320976321714736.DailyDownload.rcv `sfrbdgqS_d^y ;<CE[DCDCBFh@hJh4`h`af^c_^^aa`^geda`_e_beadhr19<Ir?G><?14hDB1>C@?BDE[ C165\D1Bh?EED213;C5BF5B Decoded Header lines: 238051003320976321714736.DailyDownload.rcv 2E8D469C 160K kls-tstsrv:p:z:d2:238051003320976321714736:DailyDownload:transport-vault.tar:outbackserver Decode the body using our script: $ encode_char.pl -r <body.tar.enc.hdr.gz.3des.enc >body.tar.enc.hdr.gz.3des

  34. “No Worries, We're Using 3DES” (9) Example decoding: putting all the pieces together, cont The key is comprised of the transaction ID + HOSTNAME: Convert the key to hexadecimal and crop to 32 characters: The hexadecimal value of the ID always exceeds 32 characters, making the inclusion of the HOSTNAME in the key irrelevant. echo '238051003320976321714736kls-tstsrv' | encode_char.pl -X | cut -c 1-32 6061665E635F5E5E6161605E67656461 Decrypted the data using the key, and ‘des’ from the ancient libdes package: des -d -3 -h -k 6061665E635F5E5E6161605E67656461 body.tar.enc.hdr.gz.3des body.tar.enc.hdr.gz Decompress the file: gzip -d body.tar.enc.hdr.gz Remove a redundant header at the end of the file: perl -ne 'print $last; $last=$_' <body.tar.enc.hdr >body.tar.enc Decode the charmapped encoded file: encode_char.pl -r <body.tar.enc >body.tar

  35. tar -tvf body.tar -rw-rw-r-- tuser/users 0 2004-05-14 15:08 out/logs.download -rw-rw-r-- tuser/users 0 2004-05-14 15:08 out/tuserreportrundetailsvault.download -rw-rw-r-- tuser/users 0 2004-05-14 15:08 out/tuserbxmcardresourcevault.download -rw-rw-r-- tuser/users 0 2004-05-14 15:08 out/tuserbxmchannelusagevault.download -rw-rw-r-- tuser/users 571 2004-05-14 15:08 out/modulesvault.download -rw-rw-r-- tuser/users 924 2004-05-14 15:08 out/modulesattrvault.download -rw-rw-r-- tuser/users 143 2004-05-14 15:08 out/downloadmastervault.download -rw-rw-r-- tuser/users 5357 2004-05-14 15:08 out/schedulervault.download -rw-rw-r-- tuser/users 0 2004-05-14 15:08 out/dailyrvault.download -rw-r----- tuser/users 187204 2004-05-14 15:08 out/logs/seedfmgr.log -rw-r----- tuser/users 1047963 2004-05-14 15:08 out/logs/scheduler.log [snip] “No Worries, We're Using 3DES” (10) • Here's a sample payload after it had been cracked

  36. “No Worries, We're Using 3DES” (11) • The information was encrypted with 3DES prior to transfer (as claimed in the documentation) • It was then ASCII encoded with a substitution cipher • BUT... the first few bytes of the transfer contained the encryption key used to encrypt the rest of the file! • The vendor didn’t mention that... :/ (skip back)

  37. We Have Both Kinds: AES and XOR • “Convergent technologies” application (Back)(Skip)

  38. We Have Both Kinds: AES and XOR (1) • “Convergent technologies” application • Fat MS Windows client • Clients for both x86 W2K, and WinCE • Mixture of web-based and proprietary communications • AES encryption used in some places and XOR encoding used in others • Both had problems...

  39. We Have Both Kinds: AES and XOR (2) • Initially, we found the secret AES key in a DLL using traditional reversing techniques • Strings, hex dumps, static disassembly of subtle routines like: Encrypt() and Decrypt() • Umm, great key guys: “1234567890abcdef” • Later, we used FTimes in dig mode to search all files for the observed key pattern: • DigStringRegExp=\x00[0-9A-Fa-f]{16}\x00 • This often reveals additional keys, but not this time

  40. We Have Both Kinds: AES and XOR (3) • Their choice of key suggested that the total key space was severely limited: • Total key space =25616 or 2128 • 340,282,366,920,938,463,463,374,607,431,768,211,456 • Hex characters only key space = 1616 or 264 • 18,446,744,073,709,551,616 • The difference is huge! • But is the crippled key space big enough? • Are there other limiters in play?

  41. We Have Both Kinds: AES and XOR (4) • Initially, our plan was to use the recovered key and an alternate AES implementation to handle our encryption/decryption needs • But the exported encrypt/decrypt routines were too enticing, so we chose to abuse them instead • Eliminated potential implementation gotchas • Required disassembly to determine call setup • But the results were worth the effort...

  42. ... #define KEY “1234567890abcdef” typedef DWORD(__stdcall *MYROUTINE)(LPTSTR,LPTSTR,DWORD,LPTSTR); int main() { ... Handle = LoadLibrary(“Application.dll”); if (Handle != NULL) { Routine = (MYROUTINE) GetProcAddress(Handle, “Decrypt”); if (Routine != NULL) { Routine(CipherText, PlainText, CHUNK_SIZE, Key); ... We Have Both Kinds: AES and XOR (5) • Here's a pseudo C implementation that shows how we used the exported Decrypt routine

  43. We Have Both Kinds: AES and XOR (6) • Now we could encrypt and/or decrypt • Local files (e.g., user profile) • Messages transmitted to/from the server • This allowed us to craft custom files and messages • It also revealed a new problem: • The implementation used ECB instead of CBC • ECB – Electronic Code Book • CBC – Cipher Block Chaining • ECB was not a good choice in this case... (skip example)

  44. 00000000 c5 e1 c6 6f 93 4a 73 f1 5a 49 e5 27 fe d9 b6 58 |...o.Js.ZI.'...X| 00000010 e3 6f 87 24 f3 b1 ed XX XX XX XX XX XX XX XX XX |.o.$............| 00000020 ef 0b f2 b7 c9 0f b5 e9 80 de 51 9a 16 f8 82 40 |..........Q....@| 00000030 f0 60 a8 d5 35 74 84 1c e5 97 43 XX XX XX XX XX |.`..5t....C.....| 00000040 ff 5b 0c e9 07 8b d0 8e d2 fe a4 0e fb a5 ca 5d |.[.............]| 00000000 3c 43 75 73 74 6f 6d 65 72 3e 3c 41 63 63 6f 75 |<Customer><Accou| 00000010 6e 74 4e 61 6d 65 3e XX XX XX XX XX XX XX XX XX |ntName>XXXXXXXXX| 00000020 34 3c 2f 41 63 63 6f 75 6e 74 4e 61 6d 65 3e 3c |4</AccountName><| 00000030 47 72 6f 75 70 73 3e 3c 49 44 3e XX XX XX XX XX |Groups><ID>XXXXX| 00000040 3c 2f 49 44 3e 3c 4e 61 6d 65 3e 46 72 69 65 6e |</ID><Name>Frien| We Have Both Kinds: AES and XOR (7)

  45. 00000000 c4 e1 c6 6f 93 4a 73 f1 5a 49 e5 27 fe d9 b6 58 |...o.Js.ZI.'...X| 00000010 e3 6f 87 24 f3 b1 ed XX XX XX XX XX XX XX XX XX |.o.$............| 00000020 ef 0b f2 b7 c9 0f b5 e9 80 de 51 9a 16 f8 82 40 |..........Q....@| 00000030 f0 60 a8 d5 35 74 84 1c e5 97 43 XX XX XX XX XX |.`..5t....C.....| 00000040 ff 5b 0c e9 07 8b d0 8e d2 fe a4 0e fb a5 ca 5d |.[.............]| 00000000 71 f7 95 1e 19 d3 bd 5e 6d 71 e9 eb 45 18 61 84 |q......^mq..E.a.| 00000010 6e 74 4e 61 6d 65 3e XX XX XX XX XX XX XX XX XX |ntName>XXXXXXXXX| 00000020 34 3c 2f 41 63 63 6f 75 6e 74 4e 61 6d 65 3e 3c |4</AccountName><| 00000030 47 72 6f 75 70 73 3e 3c 49 44 3e XX XX XX XX XX |Groups><ID>XXXXX| 00000040 3c 2f 49 44 3e 3c 4e 61 6d 65 3e 46 72 69 65 6e |</ID><Name>Frien| We Have Both Kinds: AES and XOR (8)

  46. 00000000 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000010 e3 6f 87 24 f3 b1 ed XX XX XX XX XX XX XX XX XX |.o.$............| 00000020 ef 0b f2 b7 c9 0f b5 e9 80 de 51 9a 16 f8 82 40 |..........Q....@| 00000030 f0 60 a8 d5 35 74 84 1c e5 97 43 XX XX XX XX XX |.`..5t....C.....| 00000040 ff 5b 0c e9 07 8b d0 8e d2 fe a4 0e fb a5 ca 5d |.[.............]| 00000000 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000010 6e 74 4e 61 6d 65 3e XX XX XX XX XX XX XX XX XX |ntName>XXXXXXXXX| 00000020 34 3c 2f 41 63 63 6f 75 6e 74 4e 61 6d 65 3e 3c |4</AccountName><| 00000030 47 72 6f 75 70 73 3e 3c 49 44 3e XX XX XX XX XX |Groups><ID>XXXXX| 00000040 3c 2f 49 44 3e 3c 4e 61 6d 65 3e 46 72 69 65 6e |</ID><Name>Frien| We Have Both Kinds: AES and XOR (9)

  47. 00000000 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000010 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000020 ef 0b f2 b7 c9 0f b5 e9 80 de 51 9a 16 f8 82 40 |..........Q....@| 00000030 f0 60 a8 d5 35 74 84 1c e5 97 43 XX XX XX XX XX |.`..5t....C.....| 00000040 ff 5b 0c e9 07 8b d0 8e d2 fe a4 0e fb a5 ca 5d |.[.............]| 00000000 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000010 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000020 34 3c 2f 41 63 63 6f 75 6e 74 4e 61 6d 65 3e 3c |4</AccountName><| 00000030 47 72 6f 75 70 73 3e 3c 49 44 3e XX XX XX XX XX |Groups><ID>XXXXX| 00000040 3c 2f 49 44 3e 3c 4e 61 6d 65 3e 46 72 69 65 6e |</ID><Name>Frien| We Have Both Kinds: AES and XOR (10)

  48. 00000000 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000010 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000020 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000030 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000040 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 |................| 00000000 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000010 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000020 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000030 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| 00000040 a4 07 9e f2 f5 57 6c 70 88 83 25 f0 86 11 8a e5 |.....Wlp..%.....| We Have Both Kinds: AES and XOR (11)

  49. We Have Both Kinds: AES and XOR (12) • Did we mention that the DLL in question was deployed on both client and server components? • This implied that any vulnerabilities in the DLL could potentially be leveraged on the server • Buffer overflows in the client-server traffic • Also, this gives an attacker an advantage since s/he has plenty of time to analyze the DLL offline (Skip back)

  50. We Have Both Kinds: AES and XOR (13) • And now, the other problem: XOR • The application allowed user credentials to be “remembered” • Stored in the Registry • Obscured with a position based XOR • Trivial to reverse • What were they thinking? • We've been down that road before... <sigh>

More Related